xemacs random crashes on SMP system

[Anthony Mallet <anthony.mallet%useless-ficus.net@localhost>]

Hi.

I am observing random crashes of xemacs on an SMP/i386 system (intel quad
core CPU) since quite a long time now (I would say since January this year).

The crashes can be quite easily reproduced by running `xemacs-21.4.17 -vanilla`
(from pkgsrc), visiting a big buffer (an autoconf 'configure' script of 20k
lines is the perfect example), turning on font lock, fontifying the buffer
and searching for some text. It's not 100% guaranteed but on my system I have a
combination of configure file / search text that can crash xemacs 100% of
the time. The core dump shows evidence of memory corruption, it's always at the
same place (actually 2 differents places, it depends...) but doesn't help much.

Anyway: the problem disapears as soon as I turn all CPUs offline (all but one
;) with cpuctl. The problem also doesn't show up on another UP system with
exactly the same binaires/userland/kernel (nfs mounted). BTW the SMP system
uses binaries from an NFS partition, I don't know if this matters ...

It's been a long time now that I've been trying to investigate the origin of
the problem (trying other version of xemacs, linking with -lefence, ktracing,
trying to revert the 'cld' change that happened in the memmove() function,
trying to increase softs limits for stack size, datasize, memory use,
...). Nothing changed the behaviour. And also it never happend before this year
with the same xemacs version.

I'm now running out of ideas and starting to get pretty much conviced that the
issue is related to SMP. So if you would have some tricks that I could try, I
would be very happy to continue hunting for the beast :)

And if you can reproduce the crash with the method described above I would
feel less alone also ;)

Cheers,
Anthony